[Bug target/60563] New: FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[Bug target/60563] New: FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

            Bug ID: 60563
           Summary: FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*
           Product: gcc
           Version: 4.9.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: dominiq at lps dot ens.fr
                CC: iains at gcc dot gnu.org, ian at gcc dot gnu.org
              Host: *-apple-darwin*
            Target: *-apple-darwin*
             Build: *-apple-darwin*

Running target unix/-m32
FAIL: g++.dg/ext/sync-4.C -std=gnu++98 execution test
FAIL: g++.dg/ext/sync-4.C -std=gnu++11 execution test
FAIL: g++.dg/ext/sync-4.C -std=gnu++1y execution test

Running target unix/-m64
FAIL: g++.dg/ext/sync-4.C -std=gnu++98 (test for excess errors)
UNRESOLVED: g++.dg/ext/sync-4.C -std=gnu++98 compilation failed to produce
executable
FAIL: g++.dg/ext/sync-4.C -std=gnu++11 (test for excess errors)
UNRESOLVED: g++.dg/ext/sync-4.C -std=gnu++11 compilation failed to produce
executable
FAIL: g++.dg/ext/sync-4.C -std=gnu++1y (test for excess errors)
UNRESOLVED: g++.dg/ext/sync-4.C -std=gnu++1y compilation failed to produce
executable

There is no failure at r204199, but I see them at r204372. Since the test has
been introduced in r204360, it is likely that it has never worked on darwin.

For -m32 the run time failure is

terminate called after throwing an instance of 'int'

For -m64 the compilation aborts on

/opt/gcc/work/gcc/testsuite/g++.dg/ext/sync-4.C:12:47: error: conflicting
declaration 'typedef long int int64_t'
In file included from /opt/gcc/build_w/gcc/include-fixed/stdint.h:30:0,
                 from /opt/gcc/build_w/gcc/include/stdint.h:9,
                 from /usr/include/sys/resource.h:72,
                 from /usr/include/sys/wait.h:110,
                 from /usr/include/stdlib.h:65,
                 from /opt/gcc/work/gcc/testsuite/g++.dg/ext/sync-4.C:8:
/usr/include/sys/_types/_int64_t.h:30:20: note: previous declaration as
'typedef long long int int64_t'

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

--- Comment #1 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
If I replace everywhere int64 with int_64, the test compiles with -m64 and the
failing tests are

  t3();
  t4();
  t5();
  t6();
  t9();
  t10();
  t11();
  t12();
  t16();

only with -m64, and

  t17();
  t18();
  t19();

for both -m32 and -m64.

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[ian at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

--- Comment #2 from ian at gcc dot gnu.org <ian at gcc dot gnu.org> ---
Author: ian
Date: Wed Mar 19 03:38:21 2014
New Revision: 208674

URL: http://gcc.gnu.org/viewcvs?rev=208674&root=gcc&view=rev
Log:
    PR target/60563
    * g++.dg/ext/sync-4.C (int32_t): Remove typedef.
    (ditype): Rename typedef from int64_t.

Modified:
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/testsuite/g++.dg/ext/sync-4.C

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[ian at airs dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

Ian Lance Taylor <ian at airs dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ian at airs dot com

--- Comment #3 from Ian Lance Taylor <ian at airs dot com> ---
I patched the test to avoid using int64_t.

The test is intended to ensure that the builtin sync functions correctly throw
exceptions when compiling with -fnon-call-exceptions.  The failure here may
indicate either that that does not work on Darwin, or that GCC can not unwind
from a segmentation violation on Darwin.  I don't know which.

I marked the test to run on Darwin because cleanup-9.C runs on Darwin, and it
tests that GCC can unwind from a segmentation violation.  If cleanup-9.C passes
on your system, then I don't know why sync-4.C fails.  Somebody will have to
debug the issue.

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

Iain Sandoe <iains at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-03-20
     Ever confirmed|0                           |1

--- Comment #4 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Ian Lance Taylor from comment #3)
> I patched the test to avoid using int64_t.

thanks Ian.

> The test is intended to ensure that the builtin sync functions correctly
> throw exceptions when compiling with -fnon-call-exceptions.  The failure
> here may indicate either that that does not work on Darwin, or that GCC can
> not unwind from a segmentation violation on Darwin.  I don't know which.
> 
> I marked the test to run on Darwin because cleanup-9.C runs on Darwin, and
> it tests that GCC can unwind from a segmentation violation.  If cleanup-9.C
> passes on your system, then I don't know why sync-4.C fails.  Somebody will
> have to debug the issue.

Well - when I build the test -O0,O1,Os it passes. Will need to investigate
what's changing at O2.

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

--- Comment #5 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Created attachment 32405
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32405&action=edit
preprocessed file

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

--- Comment #6 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Created attachment 32406
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32406&action=edit
asm compiled with -std=gnu++11 -fexceptions -fnon-call-exceptions -save-temps
-fverbose-asm -dA

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

--- Comment #7 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Created attachment 32407
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=32407&action=edit
asm compiled with -std=gnu++11 -fexceptions -fnon-call-exceptions -save-temps
-fverbose-asm -dA -O1

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

--- Comment #8 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
> Well - when I build the test -O0,O1,Os it passes. 

Confirmed when running the test suite. However the following reduced test

/* { dg-do run { target hppa*-*-hpux* *-*-linux* *-*-gnu* powerpc*-*-darwin*
*-*-darwin[912]* } } */
/* { dg-require-effective-target sync_long_long_runtime } */
/* { dg-options "-fexceptions -fnon-call-exceptions -O2" } */

/* Verify that the builtin functions are correctly marked as trapping
   when using -fnon-call-exceptions.  */

#include <stdlib.h>
#include <signal.h>

typedef int ditype __attribute__ ((mode (DI)));

#define FN(IDX, RET, CALL)                    \
static RET f ## IDX (void *p) __attribute__ ((noinline));    \
static RET                            \
f ## IDX (void *p)                        \
{                                \
  return CALL;                            \
}                                \
static void                            \
t ## IDX ()                            \
{                                \
  try                                \
    {                                \
      f ## IDX(0);                        \
    }                                \
  catch (...)                            \
    {                                \
      return;                            \
    }                                \
  abort();                            \
}

FN(18, void, (__atomic_clear((ditype*)p, __ATOMIC_SEQ_CST)))

static void
handler(int)
{
  sigset_t clear;

  sigfillset (&clear);
  sigprocmask (SIG_UNBLOCK, &clear, NULL);
  throw 0;
}

int
main ()
{
  signal (SIGSEGV, handler);
  signal (SIGBUS, handler);

  t18();

  exit(0);
}

fails with -O1, but not with -O0 (-m32 and -m64).

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[iains at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

--- Comment #9 from Iain Sandoe <iains at gcc dot gnu.org> ---
sadly, this seems to be a ld64 bug - present when using uncompressed EH (which
is the default for GCC on darwin).

If I manually enable compressed EH (which is the default now for the platform)
the test passes.

Will investigate whether we can yet enable compressed EH for all of GCC; it's
possible that we still emit too many other constructs that break the ld64 atom
model.

A work-around for this issue is:
{ dg-additional-options "-fno-align-functions" { target { *-*-darwin* } } }

with a FIXME comment that this is catering for an external tool bug (as usual,
we cannot expect ld64 to be fixed on any darwin < 13).

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[howarth at bromo dot med.uc.edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

howarth at bromo dot med.uc.edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |howarth at bromo dot med.uc.edu

--- Comment #10 from howarth at bromo dot med.uc.edu ---
(In reply to Iain Sandoe from comment #9)
> sadly, this seems to be a ld64 bug - present when using uncompressed EH
> (which is the default for GCC on darwin).
> 

Have you opened a radar upstream for this ld64 bug?

[Bug target/60563] FAIL: g++.dg/ext/sync-4.C on *-apple-darwin*

[howarth at bromo dot med.uc.edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=60563

--- Comment #11 from howarth at bromo dot med.uc.edu ---
(In reply to howarth from comment #10)
> (In reply to Iain Sandoe from comment #9)
> > sadly, this seems to be a ld64 bug - present when using uncompressed EH
> > (which is the default for GCC on darwin).
> > 
> 
> Have you opened a radar upstream for this ld64 bug?

Filed radr://19802258 and uploaded a standalone test case for the darwin linker
developer to debug with.